Adaptive and Fault-Tolerant RNS-AEAD Cryptosystem with Optimized CRT Reconstruction Using ChaCha20-Poly1305 for High-Performance Secure Communication

Abstract

In the cloud computing, Internet of Things (IoT), and edge computing scenarios, modern cryptographic systems need to meet four conflicting requirements: Robust security assurances, Low compute latency, Built-in data parallelism, and Fault tolerance. This paper proposes an Adaptive Fault-Tolerant RNS-AEAD cryptosystem that meets all four requirements. The proposed framework combines an adaptive coprime modulus set \( M = \{2^k, 2^k - 1, 2^k + 1, p_r\} \), RRNS-based fault detection and correction, efficient CRT reconstruction using precomputed constants, and ChaCha20-Poly1305 AEAD encryption to achieve secure, reliable, and computationally efficient residue-domain processing. Encryption time: 0.042--0.050 ms (4–15 ASCII char plaintexts, 100 iterations, IQR filtered). Decryption time: 0.057--0.068 ms (4–15 ASCII char plaintexts, 100 iterations, IQR filtered). Experimental evaluation shows significant improvement in performance over the prevailing hybrid and conventional cryptographic approaches. The average entropy of the ciphertext is 4.985 bits/byte (62.3\% of the maximum of 8 bits/byte), uniqueness of nonces and authentication-tags are 100\%, fault-detection rate of the RRNS layer is 99.63\% (P(\text{miss}) = 1/269), and chi-square randomness testing passes at the 95\% confidence level. The IND-CPA and INT-CTXT guarantees are proven in a formal way, assuming standard conditions. The results show the proposed framework to be a practical, scalable, and theoretically sound solution for latency-critical secure communication in next-generation distributed environments.